一种汽车维修问答方法、装置、终端设备及存储介质

By combining a large language model with knowledge graphs and schema information to decompose complex fault problems into sub-tasks, filtering and pruning interfering factors, and generating optimized query statements, this system solves the problems of low query accuracy and dependence on professional language in existing technologies, and realizes an efficient and easy-to-use automotive repair question-and-answer system.

CN121210600BActive Publication Date: 2026-07-17PISTON INTELLIGENCE
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
PISTON INTELLIGENCE
Filing Date
2025-08-04
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

In existing automotive diagnostic and repair technologies, traditional natural language processing methods have low query accuracy, while knowledge graph-based methods require specialized database query languages, making them difficult to popularize among frontline repair personnel and unable to meet the accuracy requirements for complex queries.

Method used

The system employs a large language model combined with knowledge graph and schema information for enhanced retrieval generation. Fault problems are decomposed into multiple sub-tasks. Nodes and edges are filtered through a vertical knowledge base information classification model, interfering factors are pruned, declarative query statements for graph databases are generated, and the query process is optimized through prompt words.

Benefits of technology

It improves the accuracy and efficiency of fault diagnosis, reduces reliance on specialized query languages, enhances user experience, adapts to complex maintenance scenarios, and reduces computational resource consumption and query logic misleading.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121210600B_ABST
    Figure CN121210600B_ABST
Patent Text Reader

Abstract

本申请公开了一种汽车维修问答方法、装置、终端设备及存储介质，属于汽车故障诊断和维修领域，所述方法为：根据预设的第一知识图谱及其对应的Schema信息，将用户输入的故障问题分解为多个子任务；根据子任务以及Schema信息生成对应各子任务的备选路径，并根据备选路径对Schema信息进行剪枝；根据备选路径，结合预设的大语言模型，生成对应各子任务的查询语句，并根据剪枝过的Schema信息对查询语句进行优化；执行优化后的各查询语句，获取每个子任务的图数据，并将各图数据以及故障问题输入至大语言模型，获取故障问题对应的自然语言回答。通过本申请，可以在保证用户体验的同时，提高故障查询精确度。
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Mechanical intelligent question and answer maintenance method and system based on large language model and knowledge graph double-wheel driving

    CN119003709A

  • Complex task intelligent planning execution method and device, electronic equipment and storage medium

    CN119645662A